This is an implementation of the Quick Fair Queue scheduler developed
by Fabio Checconi. The same algorithm is already implemented in ipfw
in FreeBSD. Fabio had an earlier version developed on Linux, I just
cleaned it up. Thanks to Eric Dumazet for testing this under load.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We need to assume that all features will be available when registering the
netdev otherwise they are ommitted from the initial set of
dev->wanted_features. When we connect to the backed we reduce the set as
necessary due to the call to netdev_update_features() in xennet_connect().
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On error path kfree() should get pointer to memory allocated by
kmalloc() not the address of variable holding it (which is on stack).
Signed-off-by: Mariusz Kozlowski <mk@lab.zgora.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
Must declare xennet_fix_features() and xennet_set_features() before
using them.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Note: get_flags was actually broken, because it should return the
flags capped with vlan_features. This is now done implicitly by
limiting netdev->hw_features.
RX checksumming offload control is (and was) broken, as there was no way
before to say whether it's done for tagged packets.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
Issue FEAT_CHANGE notification when features are changed by
netdev_update_features(). This will allow changes made by extra constraints
on e.g. MTU change to be properly propagated like changes via ethtool.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
All callers are prepared for alloc failures anyway, so this error
can safely be boomeranged to the callers domain without super
bad consequences. ...At worst the connection might go into a state
where each RTO tries to (unsuccessfully) re-fragment with such
a mis-sized value and eventually dies.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
auth_hmacs field of struct sctp_cookie is used for store
Requested HMAC Algorithm Parameter, and each HMAC Identifier
is 2 bytes, so the length should be:
SCTP_AUTH_NUM_HMACS * sizeof(__u16) + 2
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Side effects:
- TX offloads (HW csum, scatter-gather) can be toggled now
- RX checksum is reported correctly now (it's always active)
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
This should probably get TSO available as it's basically a loopback device.
Offloads are left disabled by default - as before.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
Not tested in any way. The original code for offload setting seems broken
as it resets the features on every netback reconnect.
This will set GSO_ROBUST at device creation time (earlier than connect time).
RX checksum offload is forced on - so advertise as it is.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
There's a race (not fixed here) in smsc75xx in setting RFE_CTL that's not
properly handled via rfe_ctl_lock. Spinlock is not a good tool here, as
this has to wait for URB completion (or maybe just submission) after issuing
register write request. Otherwise, the rfe_ctl might be changed just after
spin_unlock() and device left programmed with other value.
smsc95xx has increased hard_header_len for the case of TX checksumming.
smsc75xx is fixed to advertise IP+IPV6_CSUM instead of HW_CSUM as it does
not use csum_start/csum_offset.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
dev_ethtool_get_rx_csum() won't report rx checksumming when it's not
changeable and driver is converted to hw_features and friends. Fix this.
(dev->hw_features & NETIF_F_RXCSUM) check is dropped - if the
ethtool_ops->get_rx_csum is set, then driver is not coverted, yet.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
The documentation for the USB ethernet devices suggests that
only some devices are supposed to use usb0 as the network interface
name instead of eth0. The logic used there, and documented in
Kconfig for CDC is that eth0 will be used when the mac address
is a globally assigned one, but usb0 is used for the locally
managed range that is typically used on point-to-point links.
Unfortunately, this has caused a lot of pain on the smsc95xx
device that is used on the popular pandaboard without an
EEPROM to store the MAC address, which causes the driver to
call random_ether_address().
Obviously, there should be a proper MAC addressed assigned to
the device, and discussions are ongoing about how to solve
this, but this patch at least makes sure that the default
interface naming gets a little saner and matches what the
user can expect based on the documentation, including for
new devices.
The approach taken here is to flag whether a device might be a
point-to-point link with the new FLAG_POINTTOPOINT setting in
the usbnet driver_info. A driver can set both FLAG_POINTTOPOINT
and FLAG_ETHER if it is not sure (e.g. cdc_ether), or just one
of the two. The usbnet framework only looks at the MAC address
for device naming if both flags are set, otherwise it trusts the
flag.
Signed-off-by: Arnd Bergmann <arnd.bergmann@linaro.org>
Tested-by: Andy Green <andy.green@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now we have CONFIG_ARCH_DMA_ADDR_T_64BIT. We can fix the hacky
dma_addr_t size test cleanly.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is preparation for using the generic netdev features interface,
and should have no effect in itself.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Commit 60d9f461a2 ("appletalk: remove
the BKL") added a dereference of "sk" before checking for NULL in
atalk_release().
Guard the code block completely, rather than partially, with the
NULL check.
Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With current bnx2x firmware 6.2.9, iSCSI is not supported in DCB
network, so we need to disable it. Add cnic command to disconnect
iSCSI connections and prevent future connections when DCBX negotiation
succeeds.
Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We could get hardware attention during DCB/FCoE traffic without this
fix.
Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To fix bugs when running offloaded FCoE/iSCSI traffic in multiple
Class of Service environments. In some scenarios, traffic could stop
on certain rings and eventually all traffic would stop.
Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On-stack initialization via assignment of flow structures are
expensive because GCC emits a memset() to clear the entire
structure out no matter what.
Add a helper for ipv4 output flow key setup which we can use to avoid
the memset.
Signed-off-by: David S. Miller <davem@davemloft.net>
We should reduce the number of reserved completion queues from the total
number of entries. Since the queue size is power of two, not reducing the
reserved entries, caused a double queue size, which may lead to allocation
failures in some cases.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case of allocation failure, tried to use the promiscuous QP
entry that was previously freed.
Now freeing this entry only in case we will not put it back to the list
of promiscuous entries.
Reported-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add __rcu annotations and lockdep checks.
Add const qualifiers
node_parent() and node_parent_rcu() can use
rcu_dereference_index_check()
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The IRQF_SAMPLE_RANDOM flag is marked as deprecated and will be removed.
Every input point to the kernel's entropy pool have to better document the
type of entropy source it is.
drivers/char/random.c now implements a set of interfaces that can be used for
devices to collect enviromental noise. IRQF_SAMPLE_RANDOM will be replaced
with these add_*_randomness exported functions.
Network drivers are not a good source of entropy. They use as a source of
entropy essentially a remote host. Which means that the source of entropy can
be potentially controlled by an attacker. Also, with heavy workloads the
entropy decreases due to less hardware interrupts happening thanks to irq
mitigation and NAPI.
If a system relies in its network interface as a entropy source it has a false
sense of security. Systems that don't have devices whose drivers are good
sources of entropy, should either use a hardware random number generator or
feed the kernel's entropy pool from userspace using other sources of entropy
such as EGD, video_entropyd, timer_entropyd and audio-entropyd.
Signed-off-by: Javier Martinez Canillas <martinez.javier@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current enic code only supports ASSOCIATE and DISASSOCIATE port profile
operations. This patch adds enic support for port profile
PORT_REQUEST_PREASSOCIATE_RR operation. The VIC adapter (8021qbh) is capable
of handling port profile requests done in two steps namely PREASSOCIATE_RR
and ASSOCIATE today. The motivation to support PREASSOCIATE_RR comes mainly
from its use as an optimization during VM migration ie, to do resource
reservation on destination host before resources on source host are released.
PREASSOCIATE_RR is a VDP operation and according to the latest at IEEE,
8021qbh will also need to support VDP commands.
In addition to handling the new PORT_REQUEST_PREASSOCIATE_RR operation
this patch also does the below:
- Introduces handlers for PORT_REQUEST operations
- Moves most of the port profile handling code to new files enic_pp.[ch]
- Uses new fw devcmds for port profile operations
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch does the following:
- Introduces a new macro VIC_PROVINFO_ADD_TLV
- Adds a new OS type in vic_generic_prov_os_type
- Changes some vic_provinfo* helper routine args to constants
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds wrapper routines to new port profile related fw devcmds and
removes the old ones
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces new fw devcmds for port profile handling.
These new commands are similar to the current fw commands for
port profile handling. The only difference being that the new
commands split the existing port profile handling devcmds into multiple
fw commands, giving the driver finer control over port profile operations.
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Noticed by sparse:
drivers/net/via-rhine.c:1706:16: warning: cast to restricted __be16
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the more common/verbose logging styles.
Add #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
Remove smsc911x prefixes from format strings.
Rename SMSC_WARNING to SMSC_WARN.
Remove DPRINTK macro.
Use netif_<level> in SMSC_<level> macros.
Convert NETIF_MSG_<foo> uses to lower case.
Add no_printk verification in non-debug uses.
Add pdata to SMSC_<level> uses to avoid hidden variable uses.
Convert printks to netdev_<level> as appropriate.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As the be_worker() workqueue is scheduled in be_probe() it must
be canceled unconditionally in be_shutdown().
Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
stephen hemminger
Ian Campbell
David S. Miller
Jan Engelhardt
Mariusz Kozlowski
Eric Dumazet
Michał Mirosław
Ilpo Järvinen
Wei Yongjun
Arnd Bergmann
FUJITA Tomonori
Ben Hutchings
Dmitry Kravkov
Yevgeny Petrilin
Peter Pan(潘卫平)
Javier Martinez Canillas
Roopa Prabhu
Harvey Harrison
Alexey Dobriyan
Joe Perches
Sathya Perla